codec_qwen_librivox / aishell2 /data_convert.py
ftshijt's picture
Upload folder using huggingface_hub
1497fdf verified
raw
history blame contribute delete
962 Bytes
import json
import argparse
if __name__ == "__main__":
parser = argparse.ArgumentParser("convertor for json dataformat")
parser.add_argument("src", type=str)
parser.add_argument("tgt", type=str)
args = parser.parse_args()
with open(args.src, "r", encoding="utf-8") as source_json:
target_file = open(args.tgt, "w", encoding="utf-8")
count = 0
for line in source_json:
if len(line.strip()) == 0:
continue
info = json.loads(line.strip())
# speech, other = info["inputs"].split(" ", max_split=1)
update_json = {
"instruction": info["inputs"],
"output": info["targets"],
"input": "",
}
target_file.write(json.dumps(update_json))
target_file.write("\n")
count += 1
if count % 1000 == 0:
print("process {}".format(count))